Direct transcriptional control of the Dpp target omb by the DNA binding protein Brinker

Abstract

The gradient morphogen Decapentaplegic (Dpp) organizes pattern by inducing the transcription of different target genes at distinct threshold concentrations during Drosophila development. An important, albeit indirect, mode by which Dpp controls the spatial extent of its targets is via the graded downregulation of brinker, whose product in turn negatively regulates the expression of these targets. Here we report the molecular dissection of the cis-regulatory sequences of optomotor-blind (omb), a Dpp target gene in the wing. We identify a minimal 284 bp Dpp response element and demonstrate that it is subject to Brinker (Brk) repression. Using this omb wing enhancer, we show that Brk is a sequence-specific DNA binding protein. Mutations in the high-affinity Brk binding site abolish responsiveness of this omb enhancer to Brk and also compromise the input of an unknown transcriptional activator. Our results therefore identify Brk as a novel transcription factor antagonizing Dpp signalling by directly binding target genes and repressing their expression.

Introduction

Members of the transforming growth factor-β (TGF-β) superfamily of cytokines elicit a wide array of cellular responses, including the regulation of cell division, survival and specification of developmental fates (Kingsley, 1994). The Drosophila TGF-β homologue Decapentaplegic (Dpp) has been shown to function as a morphogen, i.e. it can exert long-range patterning effects and specify positional identities in a concentration-dependent manner (Lecuit et al., 1996; Nellen et al., 1996; Podos and Ferguson, 1999). Dpp is required for multiple aspects of embryonic and adult development, such as specification of the dorsoventral axis in the embryo (Irish and Gelbart, 1987; Ferguson and Anderson, 1992; Podos and Ferguson, 1999), mesoderm and endoderm induction (Frasch, 1995; Bienz, 1997), tracheal cell migration (Vincent et al., 1997) and patterning of the adult appendages (Spencer et al., 1982; Capdevila and Guerrero, 1994; Zecca et al., 1995).

Dpp is expressed in the developing wing disc, in response to short-range signalling by Hedgehog, in a thin central stripe of cells along the anteroposterior compartment boundary (Basler and Struhl, 1994; Tabata and Kornberg, 1994). From this position, Dpp moves outwards and exerts a long-range organizing effect in both compartments. Three presumed Dpp target genes have been identified, namely vestigial (vg), optomotor-blind (omb) and spalt (sal), all of which are expressed in broad nested domains centred on the Dpp expression stripe (Kim et al., 1996; Lecuit et al., 1996; Nellen et al., 1996). Expression of these targets appears to be triggered by distinct thresholds of Dpp concentration (Lecuit et al., 1996; Nellen et al., 1996).

In the past few years, several previously unknown, well conserved components of the Dpp signalling pathway have been discovered; this has led to the following paradigm of signal transduction. The binding of Dpp to a type II and type I receptor complex [consisting primarily of the serine threonine kinases encoded by punt and thickveins (tkv); Brummel et al., 1994; Nellen et al., 1994; Penton et al., 1994; Letsou et al., 1995; Ruberte et al., 1995] leads to phosphorylation of receptor-specific Smads [such as the Mothers against dpp (Mad) gene product; Sekelsky et al., 1995]. The phosphorylated Smad then associates with a general Smad partner, such as the Medea (Med) gene product (Das et al., 1998; Hudson et al., 1998; Inoue et al., 1998; Wisotzkey et al., 1998), and the Smad complex translocates to the nucleus where it functions together with other transcription factors to regulate expression of target genes (Massague and Wotton, 2000). Drosophila Mad was the first Smad demonstrated to possess direct DNA binding activity (Kim et al., 1997). Results from work carried out on the Dpp response elements from vestigial and tinman strongly indicate that direct binding of Smad proteins to enhancer sequences is a critical feature of Dpp signal transduction and that, at least for these enhancers, Smad sites play essential roles in activating gene transcription (Kim et al., 1997; Xu et al., 1998).

The recent cloning and characterization of the brinker (brk) gene is a novel twist to the possible mode by which Dpp signalling controls target gene expression (Campbell and Tomlinson, 1999; Jazwinska et al., 1999a,b; Minami et al., 1999; Ashe et al., 2000). brk is essential for proper patterning of both the embryonic ectoderm and the wing disc, indicating that it is an integral component of the Dpp system of positional information. In the wing disc, brk is expressed in the lateral-most regions of the disc adjacent to cells that express omb. Its expression is negatively regulated by Dpp signalling and this repression requires the zinc finger protein Schnurri (Marty et al., 2000). brk mutant clones lead to outgrowths and pattern duplications in the anterior and posterior regions of the wing, indicating that loss of brk results in ectopic activation of the Dpp pathway. brk clones outside the endogenous sal and omb expression domains but within the wing pouch show strong ectopic omb expression and weak sal expression in a strictly cell-autonomous manner. Epistasis experiments using double-mutant combinations of brk with either tkv or Mad showed that Dpp signalling is not required to activate Dpp target genes in cells lacking brk function (Campbell and Tomlinson, 1999; Jazwinska et al., 1999a). However, direct activation of Dpp targets by Dpp/Mad has not been excluded, particularly in the case of vg and sal where the levels of ectopic expression are lower in brk mutant cells than in their endogenous expression domains. Although no brk orthologues have been identified to date, the fact that Brk is able to antagonize BMP signalling and induce dorsal cell fates in Xenopus embryos suggests that Brk function is conserved between arthropods and chordates (Minami et al., 1999).

Here we are concerned with the mechanisms by which Brk regulates Dpp target genes. The lack of obvious structural motifs in the Brk protein sequence has prevented a clear prediction of how Brk functions. Since Brk appears to be a nuclear protein, we explored the prospect that it might be a DNA binding protein. Even as a transcriptional regulator, however, Brk could function indirectly to repress Dpp targets, e.g. by transcriptionally repressing a general activator or by inducing the expression of a repressor of Dpp target genes. To discriminate between these possibilities, we set out to isolate the regulatory sequences of omb, which appears to be the clearest case of a Dpp target gene whose regulation is primarily under Brk control. We report here that Brk is a sequence-specific DNA binding protein that interacts strongly with a site within the minimal omb wing-specific enhancer. We also provide combined genetic and biochemical evidence that Brk binding to the omb enhancer is responsible for its repression in lateral regions of the wing imaginal disc. Our results identify Brk as a novel transcription factor antagonizing Dpp signalling by directly binding target genes and repressing their expression.

Results

Isolation of the omb imaginal disc enhancers

The omb gene was localized within a genomic walk of 340 kb (Pflugfelder et al., 1990). By mapping chromosome breakpoints and neighbouring loci, the minimal and maximal extents of the omb gene were determined to be ∼80 and 160 kb, respectively (Pflugfelder et al., 1990). We systematically scanned 100 kb of the omb locus for regulatory elements; this region comprises 45 kb of upstream sequences and the first four introns (see Figure 1A). Restriction fragments from genomic lambda phages were cloned into a lacZ reporter P-element and assayed for enhancer activity in imaginal discs upon germline transformation (see Materials and methods). Three of these constructs showed reproducible, tissue-specific enhancer activities (Figure 1), each mimicking a subset of those of the endogenous omb gene. pomb7 is located 27 kb upstream of the omb transcription unit and drives lacZ expression in the wing imaginal disc. pomb19 [a subfragment of which has been used as a marker for polar photoreceptor axons by Newsome et al. (2000)] and pomb23 comprise intronic sequences and cause lacZ transcription in the eye and leg disc, respectively (Figure 1A). Although all three enhancers are Dpp responsive (not shown), they are well separated from each other and do not show significant overlap in tissue specificity. Below we focus solely on the wing enhancer.

Fig. 1. Identification of a minimal omb wing enhancer. (A) Molecular dissection of the omb regulatory region. The omb locus, indicating genomic positions according to Pflugfelder et al. (1990), is represented below the scale bar (a segment of the region from –130 to –20 kb is shown). The omb mRNA is depicted above the scale bar. All sites used for cloning (indicated as lines going below the line representing the genomic sequence) and all endogenous HindIII sites (indicated as lines that stop at the line representing the genomic sequence) are indicated. Phages from the omb walk are represented as lines and carry their original numbers (Pflugfelder et al., 1990). Fragments used for lacZ reporter constructs to scan the region for enhancer elements are shown as filled rectangles with the names indicated above. B, E, H and S represent BamHI, EcoRI, HindIII and SalI sites, respectively. ND represents constructs whose expression pattern has not been determined. The lacZ expression patterns of the three enhancers (wing, eye and leg) are shown below the corresponding constructs. (B) Wing enhancer break-up. A restriction map of the 6 kb pomb7 enhancer fragment is shown. Constructs pomb31–pomb35 are non-overlapping reporter constructs and pomb36–pomb39 are overlapping reporter constructs that were generated using available restriction sites. The 1.5 kb pomb33 fragment is enlarged below, with the positions of the terminal deletion constructs WF13–WF18 and the 300 bp overlapping constructs (WF4–WF12) marked. pomb33, pomb37 and pomb38 led to identical expression patterns. pomb34 and pomb39 showed only a very weak expression in the wing pouch. The other derivatives in this series showed no expression. WF13 and WF14 showed expression identical to pomb33. WF15 leads to a strong expression along the dorsoventral boundary and very weak expression in the rest of the pouch. WF16 showed narrower expression than pomb33, resembling that of the spalt gene. WF17 showed a pattern similar to that of WF16, but the expression was very patchy. The 900 bp WF14 fragment is shown enlarged on the right, with, beneath it, a depiction of the internal deletion constructs derived from it. In each case the deleted sequence is replaced by a single EcoRI site. None of the internal deletions had any effect on the wild-type expression pattern.

Identification of a minimal Dpp-responsive omb wing enhancer

The 6 kb pomb7 wing enhancer was analysed further by establishing a restriction map and deriving five non-overlapping and four overlapping subfragments for in vivo assays (Figure 1B). The 1.5 kb XhoI–HindIII fragment (pomb33) showed a lacZ expression pattern indistinguishable from that of wild-type omb. To narrow down further the sequences that are required and those that are sufficient for omb wing enhancer activity, four series of pomb33 derivatives were generated. In the first series, terminal deletions were introduced from each end (WF13–WF18; Figure 1B). The second series consists of nine 300 bp subfragments, each one overlapping by 150 bp with its neighbouring fragment (WF4–WF12; Figure 1B). This redundancy was applied to avoid disrupting vital transcription factor binding sites as a result of the way in which adjacent fragments were divided. Of these, WF12 was the only fragment that could drive lacZ expression in the wing disc. In the third series, internal deletions within WF14 (which still mimics endogenous omb expression) were generated (Figure 1B, inset to the right).

Our results from the analysis of subfragments of pomb33 and derivatives of WF14 suggested to us that there might be considerable redundant regulatory information encoded within the larger fragments. We therefore focused on WF12 as the minimal enhancer that still recapitulates important aspects of omb expression in the wing pouch. Unlike omb, the WF12-driven lacZ reporter is not expressed in the prospective notum region (Figure 2A and C). However, we ascertained that WF12 is Dpp responsive by examining its activity in cells expressing Tkv^QD, a constitutively active form of Tkv (Nellen et al., 1996). We also confirmed that WF12 activity is still subject to Brk control. WF12–lacZ expression was ectopically induced, in a cell-autonomous fashion, by gain of Tkv^QD activity (Figure 2D) or by loss of brk activity (Figure 2E). Thus, the 284 bp fragment WF12 represents a long-sought-after Dpp response element within the omb gene and is an ideal tool for dissecting the Brk-mediated regulation of a Dpp target.

Fig. 2. Expression of reporter constructs in wild-type and mutant situations. (A–C) lacZ expression patterns of omb–lacZ, brk–lacZ and WF12–lacZ, respectively (WF12-lacZ expression was similar to that of omb-lacZ in the pouch, but lacked dorsal hinge expression). Since the expression levels driven by WF12 were far stronger than those of omb–lacZ and brk–lacZ, the disc shown in (C) had to be recorded at a different exposure level. (D) WF12 is ectopically active in clones of cells expressing the constitutively activated Tkv^QD receptor (marked by the absence of the green CD2 marker). The expression levels in the clones were comparable to those in the endogenous domain (not apparent here since the endogenous domain is at a different focal plane). The upregulation of WF12 activity was strictly cell autonomous, i.e. ectopic activation was seen only within the clone and was observed only in clones lying within the wing pouch. The yellow rings around the mutant clones in the merged panel are caused by the difference between the subcellular localization of the surface CD2 marker and that of the nuclear β-Gal markers. (E) WF12 is ectopically active in cells mutant for brk (marked by the absence of the green GFP marker). The levels of WF12-driven expression in the brk loss-of-function clones were commensurate with those in the centre of the discs.

Brk is a DNA binding protein

Based on its amino acid sequence, it has been proposed that Brk is a DNA binding protein containing a possible helix–turn–helix (HTH) motif (Campbell and Tomlinson, 1999; Jazwinska et al., 1999a). We tested the DNA binding properties of Brk using in vitro synthesized full-length protein in an electrophoretic mobility shift assay (EMSA). As a DNA target, we used a double-stranded oligonucleotide derived from the omb WF12 enhancer (254–282, 284–257, see below). Brk protein is able to produce a retarded protein–DNA complex, indicating binding of Brk to the DNA (Figure 3C). Since the putative HTH motif is located at the N-terminus of the protein (amino acids 44–99), we produced a series of C-terminally truncated versions of Brk (Figure 3A). These derivatives are correctly synthesized (Figure 3B) and still bind to target DNA (Figure 3C, lanes 3–5) as confirmed by EMSA. The above experiments demonstrate that Brk does indeed contain a DNA binding domain, which is located within its N-terminal 175 amino acids.

Fig. 3. Brk is a DNA binding protein. (A) Schematic representation of the Brk protein and the C-terminal deletions generated from it. The coloured rectangular inserts correspond to stretches of repetitive amino acids. Blue represents glutamine, orange serine and brown histidine/alanine. The pink rectangle represents the region of putative homeodomain (HTH) homology. The putative repression domain (PMDLSLG) and the nuclear localization signal (RRLKKK) are also indicated (Jazwinska et al., 1999a). For the deletion constructs, the approximate positions of the restriction sites used for generating the deletions are indicated. The dark green boxes at the C-termini of the truncated constructs represent additional amino acids generated by the cloning procedure. (B) SDS–PAGE of [³⁵S]methionine-labelled proteins corresponding to full-length Brk (lane 1) and the three C-terminally deleted proteins indicated in (A) (lanes 2–4). The positions of the molecular weight markers (MW) are indicated on the left. (C) EMSA analysis of in vitro synthesized Brk full-length protein (lane 2) and C-terminally deleted proteins (lane 3, 1–515; lane 4, 1–266; lane 5, 1–175) binding to a labelled double-stranded oligonucleotide derived from the omb wing enhancer. Lane 1 is the free probe and lane 6 corresponds to 4 µl of control reticulocyte lysate. On the left, the arrows indicate the positions of the corresponding retarded bands generated by the proteins bound to the probe. ns, non-specific binding.

Brk binds to a specific site in the omb wing enhancer

On the basis of the results described above, we produced in Escherichia coli a glutathione S-transferase (GST) fusion protein containing the N-terminal 263 amino acids of Brk and used it in DNase I footprinting assays to scan the WF12 omb wing enhancer for potential Brk binding sites. GST–Brk protein strongly protected a single region of 17 nucleotides (nt) from DNase I digestion (Figure 4A). This sequence is located in the 3′ part of the enhancer (nt 266–282). Another, weakly protected region located more 5′ (nt 149–173) was not investigated further (data not shown). In a complementary assay, we scanned the WF12 fragment with a series of nine overlapping double-stranded oligonucleotides for Brk binding sites using EMSA (data not shown and Materials and methods). Only oligonucleotide 254–282/284–257 showed significant Brk binding, confirming the footprinting experiment.

Fig. 4. Definition of the Brk target site. (A) DNase I footprinting analysis of the omb wing enhancer with increasing amounts of affinity-purified GST–Brk fusion protein. The left-most lane represents the standard G+A Maxam–Gilbert sequencing reaction. The GST lanes correspond to the control reactions with 2 µg of affinity-purified GST protein. The protected region with the corresponding nucleotide sequence is indicated in brackets to the right. On the left, one hypersensitive site at the 3′ border of the protected region is also indicated by an arrow marked ‘hs’. (B) Alignment of the single sequence protected by GST–Brk in the omb wing enhancer with four regions protected in the lab550 enhancer. The bottom sequence represents the consensus sequence derived for Brk DNA binding. The stars above the omb-derived sequence indicate nucleotides in the putative consensus core that, when mutated, abolished Brk binding; the dashes represent nucleotides that, when mutated, did not affect Brk binding [see (C)]. (C) Sequence of the oligonucleotides derived from the omb wing enhancer carrying single point mutations; these oligonucleotides were used in the EMSA shown in (D). The grey box is delimiting the putative consensus core for Brk binding. The mutated oligonucleotides (mut275–mut271 and mut4X) that abolish Brk binding are grouped separately. (D) EMSA of the binding of in vitro synthesized full-length Brk protein to the different oligonucleotides carrying point mutations as indicated in (C). To each double-stranded oligonucleotide (indicated at the bottom), either Brk (4 µl of reticulocyte lysate) or control reticulocyte lysate (RRL; 4 µl of reticulocyte lysate without DNA template) were added. For wild type, mut274, mut271, mut268 and mut4X, the extra lane on the right represents the control probe. The arrows on the left indicate the position of the band generated by Brk or by the free probe. In addition to the probes shown in this figure, we also tested probes of mut270 and mut276, both of which bound Brk normally.

To identify a consensus sequence for Brk target sites, we analysed another Dpp-responsive enhancer, lab550 (Tremml and Bienz, 1992; Grieder et al., 1997), using the same footprinting strategy (data not shown). All the sites protected by GST–Brk in the two enhancer elements are listed in Figure 4B. From this alignment we derived a consensus sequence, GGCGC/TC/T, for high-affinity Brk binding. To assess the importance of each nucleotide, we scanned the WF12 Brk binding site with single point mutations (Figure 4C) and tested the ability of these mutants to bind in vitro synthesized full-length Brk protein. Mutations in the five central nucleotides of the Brk site (marked with asterisks in Figure 4B) abolished Brk binding (Figure 4D), indicating that these residues form the core recognition motif and that Brk is a DNA binding protein with a high target site specificity.

The Brk binding site in WF12 is also a target of an essential transcriptional activator

Loss of Brk activity in vivo results in ectopic activation of the WF12 enhancer, so one would predict that mutations in WF12 that abolish Brk binding should also lead to ectopic enhancer activity. To test this hypothesis, we mutated the core of the Brk binding site from GGCGCC to GATATC (mut4X, Figure 4C). As expected, this change completely abolished Brk binding in EMSA analysis (Figure 4D). We introduced the same 4 bp mutation into the WF12–lacZ reporter transgene. Unexpectedly, this mutation completely abolished lacZ expression rather than expanding it. We interpret this result as an indication that the Brk site overlaps with that of an activating input. Next we introduced into WF12–lacZ single base pair mutations that interfere with Brk binding (see above), anticipating that at least some of these mutations would still allow the unknown activator to bind, resulting in an uncoupling of the two inputs.

Brk binding to the omb enhancer is essential for lateral repression

Two of the five point mutations that prevent Brk binding (see Figure 4D), mut271 and mut272, completely abolished WF12 enhancer activity as observed with the 4 bp mutation. However, mutations mut273, mut274 and mut275 still expressed lacZ, albeit only in a narrow stripe along the dorsoventral boundary (Figure 5). This pattern unravels a hitherto masked, strong input into WF12 from a dorsoventral patterning system (see Discussion). Although these mutant enhancers show a reduced extent of expression along the dorsoventral axis, all of them exhibit a clear expansion in expression along the anteroposterior axis (see Figure 5). This latter property corresponds to the behaviour expected from the loss of a functional Brk binding site in the WF12 enhancer. We interpret these observations to indicate that mut273–mut275 represent mutations that completely abolish Brk binding but only partially prevent input by the activator. We further found that mutations flanking the Brk binding site (mut269, mut270, mut276 and mut277) abolish expression of WF12–lacZ (data not shown), indicating that the binding site of the activator extends beyond that of Brk.

Fig. 5. In vivo effects of point mutations in the Brk footprint. The sequence from position 254 to 285 of the WF12 enhancer is shown. The sequence stretch protected by Brk (positions 266–282) in DNase I footprinting analysis is represented as a filled box shown above the sequence. The five nucleotides presumed to form the Brk core are shown in bold and have stars above them. The five mutations (mut271–mut275) that were introduced are represented. As can be seen from the lacZ staining obtained with each of the mutations, mut273, mut274 and mut275 result in expanded expression domains while mut271 and mut272 abolish expression.

Genetic support for the loss of Brk responsiveness

We sought to validate our assumption that mut273– mut275 exhibit an extended expression along the dorsoventral boundary due to loss of Brk-mediated repression. Both the wild-type WF12 enhancer and the mutant derivatives were examined in cells that ectopically express Brk protein from a tub>CD2>brk flip-out transgene. While Brk potently repressed the transcriptional activity of WF12 (Figure 6A), it did not repress the mutant enhancers (Figure 6B). Thus, the ability to bind Brk in vitro, the lateral repression by endogenous Brk and the responsiveness to ectopic Brk in vivo all correlate with single nucleotide exchanges in the Brk core binding site. Together, these results are taken as evidence that the wild-type omb WF12 enhancer is a direct target of Brk repression.

Fig. 6. Effect of ectopic Brk expression on the enhancer activity of wild-type and mutant WF12. (A) WF12-driven reporter gene expression is potently repressed by low levels of Brk. Using a tub>CD2>brk construct, clones were generated (marked by the absence of green CD2 staining) that express brk under the control of the tubulinα1 promoter. (B) mut274–lacZ is not repressed in clones of cells expressing Brk ectopically, irrespective of whether these clones are located at the centre or at the periphery of the disc.

Discussion

Dpp regulates numerous cellular events during development. Most prominently, gradients in Dpp activity control cell fates along the dorsoventral axis of the early embryo and along the anteroposterior axis of imaginal discs (Podos and Ferguson, 1999). All of the known responses to Dpp can be attributed to the transcriptional induction or repression of target genes. Major interest is devoted, therefore, to the mechanisms by which Dpp signalling controls gene expression. Important advances have recently been made in Drosophila by the discovery that a significant aspect of Dpp target gene control involves the repressive action of Brk, whose expression itself is regulated by Dpp (Campbell and Tomlinson, 1999; Jazwinska et al., 1999a,b; Minami et al., 1999). Here we addressed the molecular mechanism by which Brk might exert its influence on Dpp target genes. Our experiments demonstrate that Brk can (i) bind DNA in a sequence-specific manner and (ii) act directly on Dpp target genes by repressing their transcription.

The omb wing enhancer

Key to our advance was the identification of regulatory elements of omb, one of a few genes that are positively regulated by Dpp in the wing imaginal disc. omb is particularly suited for the analysis of Brk function, because it is the only known Dpp target gene whose expression along the Dpp-regulated axis appears to be controlled by Brk alone. This conclusion is derived from the finding that brk single-mutant or brk tkv double-mutant cells express omb at equivalent levels (Campbell and Tomlinson, 1999; Jazwinska et al., 1999a). Therefore, normal omb activation could be achieved by simple relief from Brk repression without invoking direct input from the Mad–Medea signalling complex. By contrast, ectopic expression of sal in brk mutant clones that additionally lack either tkv or Mad is lower than the normal level of expression of spalt in the centre of the disc (Campbell et al., 1999; Jazwinska et al., 1999a; Marty et al., 2000). Hence, for genes such as sal, a higher complexity of regulatory input is expected. To circumvent the fact that the omb locus is particularly large (Pflugfelder et al., 1990), we identified the omb wing enhancer and isolated a subfragment that was <300 bp and hence well suited for molecular and biochemical analysis.

Brk is a DNA binding protein

It has been proposed that brk encodes a nuclear protein with features of a transcriptional repressor (Campbell and Tomlinson, 1999; Jazwinska et al., 1999a). These features include several putative nuclear localization signals, potential repression domains and a stretch of amino acids in the N-terminus of Brk with a high probability of folding into a HTH structure. Within the putative recognition helix, several amino acids are identical to residues found at the equivalent position in the recognition helix of many homeodomains, suggesting a DNA binding function. Our biochemical analyses demonstrate that the N-terminal part of the Brk protein binds DNA with high affinity and selectivity. In the omb wing-pouch enhancer, we identified a single high-affinity site. The dpp-dependent enhancer of the lab gene contains four Brk binding sites and is repressed by ectopic Brk in the endoderm (data not shown). Alignment of the five high-affinity sites identified in lab and omb allowed us to derive a consensus for Brk binding: GGCGC/TC/T. In contrast to homeodomain binding sites, which are rich in AT (Gehring et al., 1994), the Brk site exhibits a high GC content, suggesting that the DNA contacting residues in these proteins differ. Structural studies will be required to elucidate the details of the DNA binding mode of Brk and to find out whether the Brk DNA binding domain folds into a three-dimensional structure similar to that of homeo- domains. In addition, structure–function analyses using reverse genetics combined with biochemical studies will be necessary to determine important regions in the protein that make Brk such a potent repressor for Dpp target genes.

Unexpected complexity at the Brk binding site

Surprisingly, reverse genetics at the omb WF12 Brk binding site did not mimic the results obtained from forward genetics. Substitution of the Brk binding site did not result in ectopic expression of WF12–lacZ, but rather in its extinction. This result can be best explained by postulating that additional proteins bind to the Brk site, proteins that must confer an activation input on WF12 enhancer activity. However, some of the single point mutations in the Brk core site did allow residual activating input to occur while still preventing Brk binding. These mutations (mut273–mut275) are key for our conclusion that direct binding is relevant for the lateral repression of WF12 activity by Brk.

What could be the nature of the positive input at the Brk site? The simplest view is to propose a Dpp signalling input at this site, e.g. direct binding of a Mad–Medea complex. However, from genetic experiments described above, we know that Dpp input is dispensable in a situation in which Brk is absent, so it should also be possible to dispense with Dpp input in the equivalent situation in which Brk is prevented from binding. We envisage two scenarios that are compatible with our results. In the first scenario, it is assumed that, in addition to Brk, two proteins bind to the core Brk site. One would be a general wing-pouch activator that can only bind to an intact Brk site. The other protein would be a margin activator that requires a GC at positions 271 and 272 for binding, but does not exhibit a requirement for positions 273–275. Hence, mutations at 271 and 272 lead to a complete extinction of expression, while mutations at 273–275 still allow activation of expression along the presumptive wing margin. In the second scenario, only a single activator is proposed. This activator would be most abundant or most active in the vicinity of the margin. While mutations at 271 and 272 would be incompatible with binding, mutations at 273–275 would only reduce, but not obliterate, binding to this activator. In the presence of these latter mutations, the WF12 enhancer would therefore only be activated at sufficient levels near the margin. For both scenarios, the activity or expression of the presumed activator is likely to be under the control of wingless (Wg) signalling. Indeed, it has been noted previously that omb expression is dependent on Wg input (Grimm and Pflugfelder, 1996). Moreover, we have observed that WF12–lacZ expression is upregulated in cells expressing a constitutively active form of Armadillo (data not shown). However, we have not been able to detect binding sites for Pangolin (van de Wetering et al., 1991; Brunner et al., 1997) or Vestigial/Scalloped (Halder et al., 1998; Simmonds et al., 1998).

Mode of action of Brk

Our discovery that an intact Brk binding site is required for the activity of the presumed activating factors opens the possibility that these factors compete with Brk for DNA binding. Thus, in a simple, yet speculative, model, Brk could antagonize expression of omb by preventing access of these transcriptional activators to DNA. Although the identity of these factors and their mode of action are still unknown, our main conclusion is that Brk represses omb transcription directly by its ability to bind to omb enhancer sequences, and not via the transcriptional control of secondary repressors. An intriguing aspect of Brk that deserves exploration is its possible role in shaping the Dpp morphogen gradient and in refining the amplitudes of target gene expression. The challenge for the future will be to determine how Brk interacts with other transcriptional factors, be it on the omb enhancer or on the cis-elements of Dpp targets, which appear to be controlled by even more complex mechanisms.

Materials and methods

Reporter constructs

Phages spanning the omb walk are described by Pflugfelder et al. (1990). Phage DNA was extracted using standard protocols and the fragments used for making initial reporter constructs were obtained based on available restriction site information (Pflugfelder et al., 1990). Inserts derived from pomb33 and all subsequent derivatives were generated by PCR. The WF series of constructs were cloned as XbaI–BamHI fragments using primers with appropriate site extensions. The terminal-most derivatives were cloned as Asp718–XbaI or XbaI–XhoI fragments. Internal deletions in WF14 were generated using primers flanking the deletions such that an EcoRI site replaced the deletion in each case. All constructs were sequenced to confirm that no unwanted mutations had been introduced. Constructs carrying point mutations in the Brk footprint of WF12 were generated using the initial primer used to generate WF12 and a second primer derived from the end of the fragment and carrying a single mismatch at the relevant position. These primers were (substitution in bold):

mut275 CGGGATCCAAGCTTGGTTGCGCCTCAAGTAGC;

mut274 CGGGATCCAAGCTTGGTGTCGCCTCAAGTAGC;

mut273 CGGGATCCAAGCTTGGTGGAGCCTCAAGTAGC;

mut272 CGGGATCCAAGCTTGGTGGCTCCTCAAGTAGC;

mut271 CGGGATCCAAGCTTGGTGGCGACTCAAGTAGC.

The lacZ reporter vector used for generating the transformant lines was pX27 (Segalat et al., 1994). It contains a minimal hsp70 promoter with a canonical TATA box.

Brk expression plasmids

The cDNA for brk (kindly provided by S.Roth) was cut with DraI and HindIII, blunted with Klenow and cloned into the mammalian expression vector pSG5 (Green et al., 1988), which had been cut with EcoRI and filled in with Klenow. All C-terminal deletions were generated from pSG-brk as follows: 1–515 was cut with EagI and Eco47III, filled in with Klenow and re-ligated, generating upon translation a protein of 534 amino acids; 1–266 was cut with BamHI and re-ligated, generating a protein of 269 amino acids; 1–175 was cut with PstI and re-ligated, generating a protein of 230 amino acids. The GST fusion construct was generated by cutting a StyI fragment from pSG-brk, which contains amino acids 19–282, blunting with Klenow and ligating it into pGEX2T (Pharmacia), cut with SmaI.

Protein production

Full-length Brk was produced in vitro using the Promega TNT T7 Coupled Reticulocyte Lysate System, according to the manufacturer’s instructions. The GST–Brk fusion protein was prepared from a fresh culture of bacteria containing pGex-brk, induced with 1 mM isopropyl-β-d-thiogalactopyranoside for 4 h. Cells were harvested by centrifugation, resuspended in MTPBS (150 mM NaCl, 16 mM Na₂HPO₄, 4 mM NaH₂PO₄ pH 7.3) containing the proteinase inhibitor cocktail (Roche) and 1 mM phenylmethylsulfonyl fluoride (PMSF) and sonicated for 2× 1 min, at 50% maximal power with a Branson sonifier. After addition of Triton X-100 to 1%, the lysate was pelleted by centrifugation and glutathione–Sepharose 4B resin was added to the supernatant. After 2 h of incubation on a rotating wheel at 4°C, the resin was washed once with MTPBS, three times with MTPBS containing 0.1% NP-40 and 0.5% Triton X-100, and once with MTPBS containing 0.5% Triton X-100 and 1% Tween-20. The fusion protein was then eluted with 10 mM reduced glutathione in 50 mM Tris–HCl pH 8, 100 mM NaCl and dialysed overnight at 4°C in 10 mM Tris–HCl pH 8, 1 mM dithiothreitol (DTT), 0.1 mM PMSF, 20% glycerol.

Protein–DNA binding assays

EMSA. Double-stranded oligonucleotides corresponding to sequences in the omb wing enhancer were end-labelled by standard Klenow fill-in and purified over a Sephadex G25 spin column (Roche). Approximately 20 000 c.p.m. of labelled probe were incubated on ice for 30 min with 4 µl of reticulocyte lysate containing either full-length Brk protein or the empty expression vector, in a buffer containing 10 mM Tris–HCl pH 7.5, 50 mM KCl, 5 mM DTT, 1 mM EDTA, 4% Ficoll and 0.5 µg poly(dI–dC). The reaction mixtures were loaded on a 5% polyacrylamide/0.5× Tris–borate–EDTA (TBE) native gel and electrophoresed for ∼2 h. The gel was dried and exposed to XAR Kodak film with an intensifying screen at –70°C.

DNase I footprinting. A subcloned 344 bp fragment of the omb wing enhancer was 3′ end-labelled by standard Klenow fill-in, purified on polyacrylamide gel and incubated with 0.5–5 µg of affinity-purified GST–Brk protein and 1 µg of poly(dI–dC) for 10 min at 20°C in a binding reaction containing 50 mM MgCl₂, 20 mM Tris–HCl pH 7.9, 1 mM DTT, 25% glycerol. Two hundred nanograms of DNase I (Roche) were added to the mixture and incubation was continued for 90 s at room temperature. The reaction was stopped by addition of 0.2 M NaCl, 30 mM EDTA, 1% SDS, 100 µg/ml yeast tRNA, extracted with phenol–chloroform and precipitated in ethanol. Pellets were resuspended in 9 µl of standard 80% formamide dye, denatured for 5 min at 95°C and run on a 6% urea–polyacrylamide gel. The gel was dried and exposed to XAR Kodak film with an intensifying screen at –70°C.

The complete sequence of pomb33 has been submitted to DDBJ/EMBL/GenBank (accession no. AF291717) and is numbered from 1 to 1474. WF14 extends from nt 590 to 1474 and WF12 from nt 1191 to 1474 of pomb33. An extended version of WF12 was used for the footprint analysis. The following oligonucleotide pairs were used to scan the entire WF12 fragment for Brk binding in EMSA (oligonucleotides are numbered according to their starting and ending nucleotide positions):

1–46 CCCTCGTTTTGATGTTGCTCGCTGGCTGCCAACGCGCTTTTGTCTT;

48–3 TTAAGACAAAAGCGCGTTGGCAGCCAGCGAGCAACATCAAAACGAG;

25–71 GCTGCCAACGCGCTTTTGTCTTAAAGCTCTGGAACTGGCTGGCTGGC;

72–28 AGCCAGCCAGCCAGTTCCAGAGCTTTAAGACAAAAGCGCGTTGGC;

55–98 GGAACTGGCTGGCTGGCTGGCTGTTGCTGTTGCTTTTGTTGCCG;

99–57 TCGGCAACAAAAGCAACAGCAACAGCCAGCCAGCCAGCCAGTT;

81–128 CTGTTGCTTTTGTTGCCGATGTTGTACCGATGTTGTTGCTGGGCATTG;

129–83 TCAATGCCCAGCAACAACATCGGTACAACATCGGCAACAAAAGCAAC;

112–155 GTTGTTGCTGGGCATTGAAGACGTTGCTTATAAACGTTGATCAG;

156–110 TCTGATCAACGTTTATAAGCAACGTCTTCAATGCCCAGCAACA;

175–222 AGTGAAGTTAGTGCTCCGCAGTCCTTTTGGCCGTTCGGTTTTTATTTT;

223–177 TAAAATAAAAACCGAACGGCCAAAAGGACTGCGGAGCACTAACTTCA;

207–253 GTTCGGTTTTTATTTTATTTGTTTTTCTCCTTGTTTTTTGTATTTC;

254–209 TGAAATACAAAAAACAAGGAGAAAAACAAATAAAATAAAAACCGA;

150–178 GATCAGAAGCCGGTGCCGCCGCCTCAGTG;

184–154 TAACTTCACTGAGGCGGCGGCACCGGCTTCT;

254–282 ATACCCGGCTACTTGAGGCGCCACCAAGC;

284–257 TAAGCTTGGTGGCGCCTCAAGTAGCCGGG.

The following point mutant oligonucleotide pairs were used to test for Brk binding in EMSA:

mut280a ATACCCGGCTACTTGAGGCGCCACCATGC;

mut280b TAAGCATGGTGGCGCCTCAAGTAGCCGGG;

mut278a ATACCCGGCTACTTGAGGCGCCACAAAGC;

mut278b TAAGCTTTGTGGCGCCTCAAGTAGCCGGG;

mut277a ATACCCGGCTACTTGAGGCGCCAACAAGC;

mut277b TAAGCTTGTTGGCGCCTCAAGTAGCCGGG;

mut276a ATACCCGGCTACTTGAGGCGCCTCCAAGC;

mut276b TAAGCTTGCAGGCGCCTCAAGTAGCCGGG;

mut275a ATACCCGGCTACTTGAGGCGCAACCAAGC;

mut275b TAAGCTTGGTTGCGCCTCAAGTAGCCGGG;

mut274a ATACCCGGCTACTTGAGGCGACACCAAGC;

mut274b TAAGCTTGGTGTCGCCTCAAGTAGCCGGG;

mut273a ATACCCGGCTACTTGAGGCTCCACCAAGC;

mut273b TAAGCTTGGTGGAGCCTCAAGTAGCCGGG;

mut272a ATACCCGGCTACTTGAGGAGCCACCAAGC;

mut272b TAAGCTTGGTGGCTCCTCAAGTAGCCGGG;

mut271a ATACCCGGCTACTTGAGTCGCCACCAAGC;

mut271b TAAGCTTGGTGGCGACTCAAGTAGCCGGG;

mut270a ATACCCGGCTACTTGATGCGCCACCAAGC;

mut270b TAAGCTTGGTGGCGCATCAAGTAGCCGGG;

mut269a ATACCCGGCTACTTGCGGCGCCACCAAGC;

mut269b TAAGCTTGGTGGCGCCGCAAGTAGCCGGG;

mut268a ATACCCGGCTACTTTAGGCGCCACCAAGC;

mut268b TAAGCTTGGTGGCGCCTAAAGTAGCCGGG.

Clonal analysis and histochemistry

Tkv^QD-expressing clones were generated as described in Nellen et al. (1996) and brk mutant clones as described in Campbell and Tomlinson (1999). tub>CD2>brk clones were induced 48 h before dissection by giving a 30 min heat shock at 36°C. Imaginal discs were fixed and stained using standard protocols. Antibodies used were rabbit polyclonal anti-β-galactosidase, mouse monoclonal anti-CD2 (Serotec), rabbit polyclonal antibody against green fluorescent protein (GFP; Clontech), mouse monoclonal anti-β-galactosidase and Alexa fluorescent secondary antibodies. β–galactosidase activity was detected by fixing third instar larval discs and subjecting them to a standard X-gal colour reaction for 2 h at 37°C. In the case of pomb7, pomb19 and WF12 reporters, the staining reaction was stopped after 30 min.